Geology Reference
In-Depth Information
Neuromorphic Control Algorithm
Smart Neuromorphic Control Algorithm
As previously stated, the neuromorphic control-
ler is introduced and proved the applicability in
control engineering by Lucas et al. (2004). Neuro-
morphic control strategy is designed by mimicking
the minimum characteristics of brain emotional
processing. The
MO
is actually the control output
generated by the neuromorphic controller. In the
control application of the neuromorphic control
algorithm, the designer should determine a proper
structure for the
SI
and the
ES
purposefully. In
this article, the sensory input and emotional signal
are defined as follows
In control applications, the
ES
and
SI
should be
appropriately defined such that the control system
has the best performance. In this section, the pre-
vious form of
ES
function is redesigned using the
proportional-integral-derivative (PID) controller
(Mehrabian et al. 2006), i.e., the neuromorphic
control algorithm is used for tuning the PID
controller to improve the control performance of
seismically excited civil structures as shown in
Figure 3. The modified functions of the sensory
input and emotional signal are given by
SI
=
w y
+
w u
,
(8)
1
d
2
SI
=
w y
+
w u
,
(6)
1
d
2
d
dt
y
,
(9)
∫
∫
ES K y
=
+
K
y dt K
+
+
w
u dt
∫
P d
I
d
D
d
4
ES
=
w y
+
w
u dt
,
(7)
3
d
4
, , , and
are the weight factors
defining the relative importance that has to be
given to either signal;
y
d
is the drift response of
large smart civil structure; and
u
is the output of
the neuromorphic controller. By Equations (6)
and (7), the motivation of the suggested control-
ler and the emotional evaluation for the relation-
ship between motivation (i.e., dynamic responses
of structures) and the control action (i.e., MR
damper force) are defined. The weight factors are
determined via trial-and-errors.
To improve the performance of the proposed
neuromorphic control algorithm in Equations (6)
and (7) for vibration control of large smart civil
structures, a conventional controller proportional-
integral-derivative (PID) is integrated with the
BEL control algorithm through constructing the
part of the ES using the PID, as described in
later section. By defining new ES function, the
appropriate degree between sensory input and
controller output are evaluated in different way.
where
w w w
w
where
K K
, , and
are the weight factors
of the PID controller whose parameters are also
determined via trial-and-errors. Although the
parameters of the PID controller can be optimized
via an optimization procedure (e.g., genetic algo-
rithms), it is beyond the scope of the present re-
search. However, in near future, the authors tend
to optimize parameters used for the proposed
control algorithms. Note that the performance of
the PID controller might be sensitive to the selec-
tion of the associated weight factors: the PID
controller should be re-designed for different
structures via trial and errors.
The neuromorphic-PID control algorithm
should be modified to operate the MR dampers
for smart structure applications. It is because that
the neuromorphic-PID control algorithm gener-
ates control force signals while current or voltage
signals are required to operate of the MR damp-
ers. Thus, a conversion component that converts
the control force signals into current or voltage
K
1
2
3
4
P
I
D
Search WWH ::
Custom Search